The Markets are Nocturnal, but Not This Year
Revisiting an old observation…
Most of the stock market’s volatility comes in the daytime (open-to-close), but most of the market’s gains come overnight (close-to-open).
[growth of $1, logarithmically-scaled]
The graph above shows two hypothetical traders. The first (grey) is only long the SPY from each day’s open to close (daytime), and the second (red) from each day’s close to the following open (overnight).
Note that (a) the daytime session is about 60% more volatile, and (b) all of the SPY’s long-term gains since 1993 have come overnight and most of its long-term losses during the day.
That’s not the case this year, with all the market’s gains coming in the daytime session and both sessions being about equally volatile. Same chart, YTD…
[growth of $1, linearly-scaled]
Significance? Not sure.
In my past tests, I’ve never found a strong use for the daytime versus overnight relationship, other than polite dinner conversation.
I was digging back into the subject, and thought an updated chart was in order. As always, more to follow.
Happy Trading,
ms
. . . . .
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Stock Market Mechanics | 6 Comments
I think a lot of this has to do with European markets. What about doing a study of going long SPY after European close?
I think thats a very valid point because that was the driving force for what happened during the overnight session. I would be interested in seeing that chart as well!
Thanks,
This is really interesting.. I wonder if it’s a function of the vol of SPY ETF vs. SPX. If you look at the data for SPX it is incredibly stable over time (i.e. no strategy to make excess returns).
Also, I couldn’t replicate the first chart that you showed. I only get to around 5x by 5/8/12. Are you counting dividends, etc.?
Hello Jimmy – yes, the data is dividend-adjusted.
I’ve run alternatives to make sure this isn’t just a weird data anomaly related to the timing of dividends (i.e. removing dividends all together or manually shifting dividends to the daytime session), but the observation still holds.
michael
Got it, thank you. I ran the numbers ex. div for QQQ too and the effects are much stronger (though it seems that the outperformance is less strong after the tech bubble (same as your analysis) — perhaps a liquidity issue as volume in SPY and QQQ has increased over time?). I’m wondering why you think the strategy itself isn’t worth more than just dinner conversation?